top of page

The Science of Reward: A New Path to Overcoming Depression


The results of the study show that nEV (neural expected value) and nPE (neural prediction error) are important indicators for understanding depression outcomes, regardless of whether the patient has received treatment. This suggests that variables linked to reinforcement learning may be useful tools in developing a personalized approach to the care of people with depression, helping to identify who is most likely to improve and which interventions may be most appropriate.


Major Depressive Disorder (MDD) is a serious psychiatric condition that affects approximately 7% of the population each year and impacts more than 17 million people in the United States alone. The condition places an enormous personal and financial burden on individuals and society.


Unfortunately, less than 40% of people treated with standard interventions for MDD achieve symptom remission within three months.


These limited results have spurred research to identify factors that may influence the trajectories and outcomes of depression, seeking new ways to improve treatment.

Based on recent data, we know that depression can be characterized by atypical responses to rewards and losses. This study sought to explore whether variables associated with these processes could predict disease outcomes in people diagnosed with depression at baseline.


Some participants were followed throughout the natural course of the disease, while others underwent cognitive behavioral therapy (CBT).


The methodological approach of the study was based on reinforcement learning models, which allow us to analyze how the brain processes and learns from rewards and losses.

Previous studies using functional neuroimaging techniques have indicated that processes related to reward anticipation and reinforcement learning frequently activate specific brain areas, such as the striatum and insula, in individuals with and without psychiatric disorders.


In addition, there is evidence that the more severe the symptoms of depression, the greater the changes in the behavioral and neural components of reward learning. For example, anhedonia, the loss of interest in pleasurable activities, has been associated with a reduced ability to learn from rewards.


Anxious arousal has been linked to negative prediction errors, while negative affect appears to exacerbate attention to negative outcomes during loss learning.


Previous studies have also suggested that symptoms of depression may influence different aspects of reinforcement learning.


Previous research has shown that computational variables such as expected value (the anticipated value of a choice) and prediction error (the difference between the expected value and the actual outcome of a choice) are linked to specific depressive symptoms.


This knowledge motivated researchers to investigate whether these variables could predict the outcomes of depression, depending on the symptom profile presented by each individual.

In the context of this study, the impact of treatment was also considered. Among the participants, some underwent a course of standard cognitive-behavioral therapy, while others were simply monitored during the natural course of the disease without specific intervention.


It is known that a considerable proportion of adults with depression can achieve spontaneous remission within a period of 3 months (23%) or 1 year (53%).


Identifying who is most likely to recover naturally can help allocate therapeutic resources more efficiently. Thus, people with a higher chance of spontaneous remission could receive only monitoring, while those with a lower chance could be referred for more intensive interventions.


In this study, researchers from Virginia Tech, USA, used a technique called support vector machines to assess whether brain responses measured by fMRI (functional magnetic resonance imaging) associated with neural prediction error (nPE) and neural expected value (nEV) could predict remission of depression.


Neural prediction error (nPE) and neural expected value (nEV) are concepts related to reinforcement learning and how the brain processes rewards and losses.


nEV represents the value that the brain anticipates obtaining from a choice or action, based on previous experiences, it is like an expectation of what might happen.


nPE reflects the difference between this expectation (nEV) and the actual outcome of the choice. When the outcome is better or worse than expected, nPE signals the brain to adjust future expectations, helping with learning and decision-making.


In people with depression, these processes can be altered, impacting the way they evaluate rewards or deal with losses.

They also explored whether these predictions were affected by the type of treatment (naturalistic or cognitive behavioral therapy) or by the symptoms presented.


The study included 55 participants diagnosed with depression at baseline, of whom 36 completed a standard course of cognitive behavioral therapy and 19 were simply followed through the natural course of the illness.


All participants were assessed again at a follow-up visit to determine whether they had achieved remission.


The results showed that both nPE and nEV were significant predictors of depression remission, with nEV being a stronger predictor.


Interestingly, treatment status (whether or not the participant had undergone cognitive behavioral therapy) did not significantly influence the accuracy of these predictions.

However, a significant interaction between nEV and anhedonia was observed: individuals with higher levels of anhedonia showed a greater influence of nEV on predictions of remission. Other symptoms, such as negative affect or anxious arousal, did not have a significant impact on these relationships.


It is important to note that although the sample size is comparable to that of other studies, it limits the ability to generalize the results. To address this limitation, the researchers used two standard methods to validate the models: data splitting into 90% for training and 10% for testing, as well as bootstrap sampling.


In conclusion, the study confirmed that nEV and nEV are useful biobehavioral signals for predicting depression outcomes, regardless of the type of follow-up received.


Of the two, nEV proved to be the stronger predictor. These reinforcement learning variables may be valuable tools within a personalized medicine approach to treating depression.



READ MORE:


Reinforcement learning processes as forecasters of depression remission

Vansh Bansal, Katherine L. McCurry, Jonathan Lisinski, Dong-Youl Kim, Shivani Goyal, John M. Wang, Jacob Lee, Vanessa M. Brown, Stephen M. LaConte, Brooks Casas, and Pearl H. Chiu 

J Affect Disord. 2025 Jan 1:368:829-837.

doi: 10.1016/j.jad.2024.09.066


Abstract:


Aspects of reinforcement learning have been associated with specific depression symptoms and may inform the course of depressive illness. We applied support vector machines to investigate whether blood‑oxygen-level dependent (BOLD) responses linked with neural prediction error (nPE) and neural expected value (nEV) from a probabilistic learning task could forecast depression remission. We investigated whether predictions were moderated by treatment use or symptoms. Participants included 55 individuals (n = 39 female) with a depression diagnosis at baseline; 36 of these individuals completed standard cognitive behavioral therapy and 19 were followed during naturalistic course of illness. All participants were assessed for depression diagnosis at a follow-up visit. Both nPE and nEV classifiers forecasted remission significantly better than null classifiers. The nEV classifier performed significantly better than the nPE classifier. We found no main or interaction effects of treatment status on nPE or nEV accuracy. We found a significant interaction between nPE-forecasted remission status and anhedonia, but not for negative affect or anxious arousal, when controlling for nEV-forecasted remission status. Our sample size, while comparable to that of other studies, limits options for maximizing and evaluating model performance. We addressed this with two standard methods for optimizing model performance (90:10 train and test scheme and bootstrapped sampling). 

Results support nEV and nPE as relevant biobehavioral signals for understanding depression outcome independent of treatment status, with nEV being stronger than nPE as a predictor of remission. Reinforcement learning variables may be useful components of an individualized medicine framework for depression healthcare.




コメント


© 2024 by Lidiane Garcia

bottom of page